CrisisNet — Multi-Agent Crisis Coordination

CrisisNet is a multi-agent AI system built for food and energy security decision support. When a crisis hits — drought, energy spike, supply chain collapse, or market shock — analysts across agriculture, logistics, energy, and finance each build their own picture separately. By the time the reports reach a decision-maker, the window for early intervention has closed. CrisisNet solves this by deploying five specialised AI agents simultaneously: a Farmer agent analysing crop yields and climate risk, a Logistics agent mapping supply routes and bottlenecks, an Energy agent tracking grid load and fuel prices, a Market agent monitoring demand signals and speculation, and a Regulator agent that synthesises all four into a final policy recommendation. Each agent queries a domain-specific vector knowledge base (ChromaDB RAG) built from real FAO reports, USDA stock data, and regional agricultural statistics before calling an open-source LLM via Featherless AI. Agents coordinate through Band AI's platform — every finding is published to a shared Band Room with a full audit trail. A coordinator automatically detects conflicts between agent signals — for example, when market price recommendations contradict supply shortage data. The Regulator agent then issues a structured policy decision: emergency imports, reserve releases, subsidy plans, price controls — with a calibrated confidence score. If confidence falls below 70% or critical thresholds are breached, the system escalates to a human with a clear explanation. The output is a professional PDF report ready to share with a minister — crop forecasts by region, logistics route plans with costs, price recommendations per commodity, and subsidy proposals with budget estimates. Built on FastAPI, Band AI, Featherless AI, ChromaDB, and React.